[gmx-users] Structural features for LINCS application

[Francesco Oteri]

Il 16/03/2012 01:38, Mark Abraham ha scritto:
> On 16/03/2012 6:02 AM, Francesco Oteri wrote:
>> Dear gromacs users,
>> I am trying to simulate a protein (containing FeS cluster and a 
>> complex metal active site) using virtual site.
>> I've to face a problem with LINCS. In particular, if I constrain only 
>> h-bonds without using virtual site,
>> simulation runs fine but constraining "all-bonds", simulation crashes 
>> after a lot of LINCS warning like:
>>
>> Step 356, time 0.712 (ps)  LINCS WARNING
>> relative constraint deviation after LINCS:
>> rms 0.408533, max 8.159325 (between atoms 2750 and 2754)
>> bonds that rotated more than 30 degrees:
>>  atom 1 atom 2  angle  previous, current, constraint length
>>    2750   2754   90.0    0.1365   1.2502      0.1365
>>
>> In both cases simulation conditions are the same. The bonds causing 
>> problem belongs to the active site.
>> I am wondering if there are structural features imparing the use of 
>> all-bonds constraints in LINCS.
>> A second question is, how can I run MD with virtual site without 
>> all-bonds options.
>
> Coupled constraints, such as you might have in a cluster, can be 
> delicate. Equilibrating with a quite small time step can be necessary.
>
> Mark
Hi Mark,
I've run 100ps at 0.5fs, and 1fs...but when I switch to 2fs it crashes.
I've read in the manual and it seem that triangular structures might 
introduce problem, but I dont have such a structure in my protein.
So, there are other coupled constraint that may cause problem in LINCS? 
Is it possible that the force constants can cause problem?
Or anyway, is it possible using virtual site constraining only h-bonds?

Francesco